Weekly status #331

tomp · January 22, 2024, 9:54am

Weekly status for the week of 15th January to 21st January.

Introduction

The highlight of the past week was the release of pylxd 2.3.2. This is the first release of the python SDK client package since September 2022 and contains lots of fixes. For more information please see the release notes.

In other SDK client related news, the Go SDK client has been released under Apache-2.0 (the same as pylxd is).

The images: remote has been been removed from the lxc CLI client as the LinuxContainers.org image server will shortly no longer be available for LXD users. Please see An update on the licence change and community image server for more information.

LXD

Improvements:

Back in June 2023 with the release of LXD 5.15 a new feature called CSM (Compatibiltiy Support Module) mode was introduced which allowed booting non-UEFI VMs to aid in running legacy workloads. This was achieved by using the EDK2 firmware’s CSM mode to boot Seabios to run the guest. As of December 2023 the CSM support in the upstream EDK2 firmware has been removed and so to ensure that we can continue to use EDK2 in the future we have taken the decision to switch to booting Seabios directly when security.csm mode is enabled. This change has also allowed us to resolve some issues we were seeing with booting some Linux OSes due to a combination of their grub and kernel having EDD enabled.
The /1.0 API endpoint now includes an indication as to the instance types that the host supports. E.g.

lxc query /1.0 | jq '.environment | {server_name, instance_types}'
{
  "server_name": "server-1",
  "instance_types": [
    "virtual-machine",
    "container"
  ]
}

Introduces new server configuration keys to provide the SSL CA and client key pair to access the OVN databases (from Incus). The new configuration keys are network.ovn.ca_cert, network.ovn.client_cert and network.ovn.client_key.
Stopped instances are now included in the LXD metrics API endpoint.
Stop using --checksum option for rsync filesystem refreshes. This speeds up refreshes by avoiding needing to read files off the disks that have not been modified on either side (according to their modified time/size). This was originally added because rsync version 3.1.1 didn’t have support for checking the files modification times on nanoseconds granularity. This has since changed and is no longer needed.
The deprecated instance level limits.network.priority setting has now been removed. It only worked with cgroupv1 and the same functionality can now be specified on a per NIC device basis using the limits.priority setting.

Bug fixes:

Fixed a regression in server side when deciding whether to use instance server side move mode.
Fixed a bug where a stalled file transfer from an instance would hold open the exclusive lock for that instance preventing snapshots from being taken for a long time. File transfer connections now enable TCP keepalives to detect stalled connections and release the lock sooner.
Fixed a bug where snapshots.pattern for custom volumes was not respected when creating a manual snapshot without specifying a new snapshot name.
Fixed a bug that prevented pushing a file to a stopped unprivileged container which has an additional idmapping set via raw.idmap when not using idmapped mounts.
Detect when a live migration of an instance with custom volumes attached is requested and fail cleanly early on rather than failing abruptly mid-way through the migration.
Fixed bug where to restore a custom storage volume from a snapshot fails on Ceph RBD backed volumes if either xfs or btrfs is selected for block.filesystem.

All changes

The items listed below is all of the work which happened over the past week and which will be included in the next release.

LXD

LXD Charm

Distribution work

This section is used to track the work done in downstream Linux distributions to ship the latest LXD as well as work to get various software to work properly inside containers.

Ubuntu

Nothing to report this week.